“Ultimately, the size of the balance sheet will be determined by a number of factors, including demand for nonreserve liabilities, such as currency (which has been rising), and, importantly, the quantity of reserves necessary to remain reliably on the flat portion of the reserve demand curve.” Vice Chairman for Supervision Randal K. Quarles, February 22, 2019.
Prior to the Lehman failure in 2008, the Federal Reserve controlled the federal funds rate through open market operations that added to or subtracted from the excess reserves that banks held at the Fed. Because excess reserves typically were only a few billion dollars, the funds rate was very sensitive to small changes in the quantity of reserves in the system.
The Fed’s response to Lehman and its aftermath included large-scale asset purchases that led to a thousand-fold increase in excess reserves. Consequently, since 2008, small open-market operations of a few billion dollars no longer alter the federal funds rate. Instead, the Fed introduced administered rates to change its policy stance. The most important of these—the interest rate that the Fed now pays on excess reserves (IOER)—sets a floor below which banks will not lend to other counterparties (since an overnight loan to the Fed is the safest rate available).
The first test of the new mechanism for interest rate control occurred in late 2015, when the FOMC decided that it was time to start tightening policy. Excess reserves were still above $2.5 trillion, so (as we describe in earlier posts here and here), Fed policymakers raised the IOER rate to drive up all interest rates. As a supplementary tool, the Fed also supplied overnight repurchase agreements to a wide range of banks and nonbanks at an interest rate that was initially 25 basis points below the IOER (this is the ON RRP rate). The purpose of this these two administered rates is to keep the overnight interbank rate inside of the Fed’s 25-basis-point target range.
And, until very recently, the Fed’s ability to control the federal funds rate seemed well in hand. To see this, have a look at the following chart. First, the gray area is the 25-basis-point target range for the federal funds rate. The IOER is the red line and the daily effective federal funds rate (a volume-weighted measure of the transactions rates) is the blue line. The dashed lines show the range of daily trading in the federal funds market (marking the 1st percentile and 99th percentile of the distribution of trades). For the entirety of the period we show, the effective federal funds rate (blue line) remains within the target (gray) range. Furthermore, until recently, the effective federal funds rate (blue line) is consistently below the IOER (red line). And, at least until 2018, the dashed lines suggest that the vast majority of transactions in the federal funds market are within the target range.
Various overnight interest rates (daily), March 2016-June 2019
Sources: FRED and the Federal Reserve Bank of New York.
Starting in March 2018, things change. First, the downward spikes in the federal funds rate at the end of each month disappeared. (These spikes appear to arise from the fact that foreign banks are quite active in the fed funds market, but their willingness to borrow fed funds to finance the purchase of additional assets has costs in the form of month-end capital requirements.) Second, throughout the second half of 2018 and into 2019, the effective federal funds rate (blue line) gets closer and closer to the IOER (red line), until it finally crosses in March.
This brings up an intriguing question: With more than $1.3 trillion of excess reserves in the system, why would a bank be willing to pay more to borrow reserves than it can receive in interest from the Fed? Federal Reserve experts have suggested a number of possible technical reasons (see here and here). However, the simplest answer is that, even with this very high level of reserves, the market has shifted to a part of the demand curve for excess reserves that is very mildly upward sloping (as opposed to the flat portion of the curve that prevailed since 2008).
We view the following chart as a stylized version of what may be going on. The Federal Reserve has halved excess reserves from their peak of $2.7 trillion in 2014. As the supply of reserves shifts to the left, the effective federal funds rate rises from A (below IOER), to B (equal to IOER), and finally to C (above IOER). In a recent paper, James Hamilton notes additional evidence in favor of the view that reserve supply has contracted, driving the federal funds rate up and the quantity in the market down.
Evolution of the Market for Reserves
Notes: RD reserve demand. RS reserve supply. The shaded area in the figures denotes the federal funds rate target range.
So, banks certainly appear to want to hold massive excess reserves compared to the pre-Lehman norm. Why? One reason is that reserves now pay interest, whereas the IOER rate prior to October 2008 was zero. Knowing this, an analysis by Fed economists in early 2017 assumed that by 2022, reserve demand would stabilize at roughly $100 billion. However, just two years later, in March 2019, a Fed survey concluded that banks were likely to want well over $1 trillion in reserves. What accounts for this gap?
The answer is liquidity regulations. Today, banks must hold significant liquid assets to back various sorts of short-term liabilities. The details of the Liquidity Coverage Ratio (LCR) are complex, but the basics are simple: banks need to hold some combination of reserves and U.S. Treasury securities to guard against deposit outflows in times of stress. That is, prior to going to the Fed to borrow, these new regulations envision that banks will use the liquid assets they have on hand to meet withdrawals.
In practice, it turns out that banks prefer to hold reserves than securities to insure against the possibility of outflows. There are several reasons for this. First, if securities―even U.S. Treasuries―are sold quickly, it can drive prices down (something that banks’ own liquidity stress tests may assume). Second, everyone finds out when someone is selling securities under stress. If a bank uses reserves to meet withdrawals, only the Fed knows. The mix of liquidity considerations and the stigma from large Treasury sales makes reserves very attractive.
Returning to the first chart, note that the relationship between the IOER and the top of the target range (the gray area) has been changing. Originally at the upper end of the range, the Fed has adjusted the IOER down in three steps by 5 basis points each time, so that now it is 15 basis points lower. These technical adjustments are an attempt to retain control of interest rates as the supply of excess reserves declines.
The question is whether “technical adjustments” will be sufficient to maintain precise interest rate control. Unfortunately, a return to pre-crisis operations using small open-market operations remains infeasible given the Fed’s desire to remain near the flat part of the reserve demand curve (see the opening quote from Vice Chair Quarles). While a reserve supply change of $1 billion in 2007 would move the funds rate substantially, today, a change of $10 billion or even $100 billion may produce little or no change in the funds rate: while the demand for reserves may now be upward sloping, that slope is still quite gentle. We know this because the U.S. Treasury general account can move by $200 billion from one week to the next—adding or draining reserves from the banks--without having a large impact on the effective federal funds rate.
Having promised to keep the supply of reserves ample and to use administered rates (like the IOER) to ensure interest rate control, does the Fed need to add further to its set of tools?
Andolfatto and Ihrig recently proposed that the Fed create a repo facility, standing ready to provide cash in exchange for Treasury securities at a rate that is some premium over the IOER (see here and here). They note that such a facility would both create a ceiling on the repo rate and allow banks to economize on reserve balances since those with qualifying collateral would always have access to the Fed in the event of sudden withdrawals. This approach also reduces stigma: unlike market transactions, where word of mouth identifies sellers relatively quickly, the Fed only discloses individual transactions with a roughly two-year lag (see here).
To see where this new repo rate would fit into the existing framework, note that the Fed currently has three administered rates: (1) the discount lending rate (currently 3%); (2) the IOER (2.35%); and (3) the overnight-reverse-repurchase agreement rate (ON RRP) (2.25%). The first two are the rates for banks to borrow or deposit reserves. Because a wide set of nonbanks can engage in repurchase agreements with the Fed, the ON RRP rate sets a floor on market repo rates. (The federal funds rate target range is simply the objective for a market rate, not a rate administered by the Fed.)
Looking at this configuration of interest rates, the addition of a standing repo rate is the next logical step. In the same way that the discount rate and the IOER establish a corridor for interest rates on reserves, a repo and reverse-repo facility would create a floor and ceiling for the repo rate. (We assume that, like the reverse repo facility, the new repo facility would be available to nonbanks as well as banks.) As Andolfatto and Ihrig note, such a corridor system can eliminate upward spikes in rates should institutions and markets come under stress.
While we recognize the desire to improve interest rate control, the introduction of a standing repo facility poses significant challenges. To protect the public finances, Fed lending requires overcollateralization. That is, when a bank obtains a discount loan, the market value of the securities it posts as collateral exceeds the value of the loan. This is true even for U.S. Treasury Bills, where the haircut is 1%―that is, providing $100 in market value T-bills can obtain a $99 loan. If the same were true here, then this would make securities relatively more costly as a way to meet the LCR requirement, creating a new bias toward reserves.
A bigger problem is that the repo facility would not just cap the repo rate, but together with the haircut schedule, it would effectively put a floor on the price of the U.S. Treasury collateral. Why would anyone sell a Treasury for less than the price implied by the Fed’s standing repo rate? Suppose, for example, that (since banks can hold Treasurys of any maturity to meet their LCR) the facility would accept Treasurys of all maturities at haircuts that rise with the maturity. In practice, this would put a ceiling on the entire yield curve. That is, the repo facility would become a way of controlling the entire Treasury term structure! To avoid that outcome, the Fed would need to accept only short-term Treasurys at the standing repo facility.
And, this may not be the end of the story. Once the Fed creates a U.S. Treasury repo facility, the next time the system is under stress, there will be significant pressure to broaden acceptable collateral (albeit at varying haircuts). That is, as is the case with discount lending, a bank would be able to take virtually any collateral to the Fed and obtain reserves. In fact, Selgin proposes that we simply go there from the start.
A broad standing repo facility bears a strong resemblance to Mervyn King’s proposal for a pawnbroker for all seasons. When we wrote about the pawnbroker idea several years ago (see here), we noted the risks that the haircut schedule could become a mechanism for allocating credit in the economy. Given that the Fed’s standing repo facility would likely have counterparties beyond just banks, this risk seems much greater here.
Moving forward, what should the FOMC do? The secular decline of interbank lending suggests that the days of targeting the federal funds rate are numbered. In our view, the Fed should prepare to target an interest rate arising from the short-term collateralized market—where most funding transactions now occur. Targeting the secured overnight funding rate (SOFR) also would help speed the ongoing transition from LIBOR.
It is important to structure the interest rate control system to avoid some of the pitfalls we just discussed. We suggest that the design consider the following four points. First, to avoid influencing the entire yield curve, the target rate should rely mostly on short-term Treasurys as collateral. Second, if there is to be a standing repo facility designed to cap the rate in stress periods, set the cap significantly above the target range―at least at the level of the discount rate, if not slightly higher to reflect a bank’s regulatory costs. Admittedly, this would make interest rate control less precise, helping to sustain banks’ demand for reserves. Third, administer the current floor created by the overnight-reverse-repurchase agreement facility with quantity limits as needed to avoid the risk of absorbing significant levels of nonbank funds in periods of stress (see here). Finally, as we discuss in our post on the composition of the Fed’s balance sheet, coordinate with the Treasury debt manager to insure that there is sufficient supply to meet the market demand for short-term qualifying assets to avoid a safety premium.
In closing, we see how introducing a standing repo facility could improve interest rate control. But before doing so, it is important that Fed policymakers make sure that any changes in their operational framework are both consistent with longer-term plans for their interest rate target and that they avoid a range of potential pitfalls.